Denoising diffusion models have demonstrated outstanding results in 2D image generation, yet it remains a challenge to replicate its success in 3D shape generation. In this paper, we propose leveraging multi-view depth, which represents complex 3D shapes in a 2D data format that is easy to denoise. We pair this representation with a diffusion model, MVDD, that is capable of generating high-quality dense point clouds with 20K+ points with fine-grained details. To enforce 3D consistency in multi-view depth, we introduce an epipolar line segment attention that conditions the denoising step for a view on its neighboring views. Additionally, a depth fusion module is incorporated into diffusion steps to further ensure the alignment of depth maps. When augmented with surface reconstruction, MVDD can also produce high-quality 3D meshes. Furthermore, MVDD stands out in other tasks such as depth completion, and can serve as a 3D prior, significantly boosting many downstream tasks, such as GAN inversion. State-of-the-art results from extensive experiments demonstrate MVDD's excellent ability in 3D shape generation, depth completion, and its potential as a 3D prior for downstream tasks.
It is widely believed that sparse supervision is worse than dense supervision in the field of depth completion, but the underlying reasons for this are rarely discussed. To this end, we revisit the task of radar-camera depth completion and present a new method with sparse LiDAR supervision to outperform previous dense LiDAR supervision methods in both accuracy and speed. Specifically, when trained by sparse LiDAR supervision, depth completion models usually output depth maps containing significant stripe-like artifacts. We find that such a phenomenon is caused by the implicitly learned positional distribution pattern from sparse LiDAR supervision, termed as LiDAR Distribution Leakage (LDL) in this paper. Based on such understanding, we present a novel Disruption-Compensation radar-camera depth completion framework to address this issue. The Disruption part aims to deliberately disrupt the learning of LiDAR distribution from sparse supervision, while the Compensation part aims to leverage 3D spatial and 2D semantic information to compensate for the information loss of previous disruptions. Extensive experimental results demonstrate that by reducing the impact of LDL, our framework with sparse supervision outperforms the state-of-the-art dense supervision methods with 11.6% improvement in Mean Absolute Error (MAE)} and 1.6x speedup in Frame Per Second (FPS)}. The code is available at https://github.com/megvii-research/Sparse-Beats-Dense.
We propose SparseDC, a model for Depth Completion of Sparse and non-uniform depth inputs. Unlike previous methods focusing on completing fixed distributions on benchmark datasets (e.g., NYU with 500 points, KITTI with 64 lines), SparseDC is specifically designed to handle depth maps with poor quality in real usage. The key contributions of SparseDC are two-fold. First, we design a simple strategy, called SFFM, to improve the robustness under sparse input by explicitly filling the unstable depth features with stable image features. Second, we propose a two-branch feature embedder to predict both the precise local geometry of regions with available depth values and accurate structures in regions with no depth. The key of the embedder is an uncertainty-based fusion module called UFFM to balance the local and long-term information extracted by CNNs and ViTs. Extensive indoor and outdoor experiments demonstrate the robustness of our framework when facing sparse and non-uniform input depths. The pre-trained model and code are available at https://github.com/WHU-USI3DV/SparseDC.
We study the problem of aligning a video that captures a local portion of an environment to the 2D LiDAR scan of the entire environment. We introduce a method (VioLA) that starts with building a semantic map of the local scene from the image sequence, then extracts points at a fixed height for registering to the LiDAR map. Due to reconstruction errors or partial coverage of the camera scan, the reconstructed semantic map may not contain sufficient information for registration. To address this problem, VioLA makes use of a pre-trained text-to-image inpainting model paired with a depth completion model for filling in the missing scene content in a geometrically consistent fashion to support pose registration. We evaluate VioLA on two real-world RGB-D benchmarks, as well as a self-captured dataset of a large office scene. Notably, our proposed scene completion module improves the pose registration performance by up to 20%.
Recent works have demonstrated the importance of object completion in 3D Perception from Lidar signal. Several methods have been proposed in which modules were used to densify the point clouds produced by laser scanners, leading to better recall and more accurate results. Pursuing in that direction, we present, in this work, a counter-intuitive perspective: the widely-used full-shape completion approach actually leads to a higher error-upper bound especially for far away objects and small objects like pedestrians. Based on this observation, we introduce a visible part completion method that requires only 11.3\% of the prediction points that previous methods generate. To recover the dense representation, we propose a mesh-deformation-based method to augment the point set associated with visible foreground objects. Considering that our approach focuses only on the visible part of the foreground objects to achieve accurate 3D detection, we named our method What You See Is What You Detect (WYSIWYD). Our proposed method is thus a detector-independent model that consists of 2 parts: an Intra-Frustum Segmentation Transformer (IFST) and a Mesh Depth Completion Network(MDCNet) that predicts the foreground depth from mesh deformation. This way, our model does not require the time-consuming full-depth completion task used by most pseudo-lidar-based methods. Our experimental evaluation shows that our approach can provide up to 12.2\% performance improvements over most of the public baseline models on the KITTI and NuScenes dataset bringing the state-of-the-art to a new level. The codes will be available at \textcolor[RGB]{0,0,255}{\url{{https://github.com/Orbis36/WYSIWYD}}
Monocular depth inference is a fundamental problem for scene perception of robots. Specific robots may be equipped with a camera plus an optional depth sensor of any type and located in various scenes of different scales, whereas recent advances derived multiple individual sub-tasks. It leads to additional burdens to fine-tune models for specific robots and thereby high-cost customization in large-scale industrialization. This paper investigates a unified task of monocular depth inference, which infers high-quality depth maps from all kinds of input raw data from various robots in unseen scenes. A basic benchmark G2-MonoDepth is developed for this task, which comprises four components: (a) a unified data representation RGB+X to accommodate RGB plus raw depth with diverse scene scale/semantics, depth sparsity ([0%, 100%]) and errors (holes/noises/blurs), (b) a novel unified loss to adapt to diverse depth sparsity/errors of input raw data and diverse scales of output scenes, (c) an improved network to well propagate diverse scene scales from input to output, and (d) a data augmentation pipeline to simulate all types of real artifacts in raw depth maps for training. G2-MonoDepth is applied in three sub-tasks including depth estimation, depth completion with different sparsity, and depth enhancement in unseen scenes, and it always outperforms SOTA baselines on both real-world data and synthetic data.
Reconstructing transparent objects using affordable RGB-D cameras is a persistent challenge in robotic perception due to inconsistent appearances across views in the RGB domain and inaccurate depth readings in each single-view. We introduce a two-stage pipeline for reconstructing transparent objects tailored for mobile platforms. In the first stage, off-the-shelf monocular object segmentation and depth completion networks are leveraged to predict the depth of transparent objects, furnishing single-view shape prior. Subsequently, we propose Epipolar-guided Optical Flow (EOF) to fuse several single-view shape priors from the first stage to a cross-view consistent 3D reconstruction given camera poses estimated from opaque part of the scene. Our key innovation lies in EOF which employs boundary-sensitive sampling and epipolar-line constraints into optical flow to accurately establish 2D correspondences across multiple views on transparent objects. Quantitative evaluations demonstrate that our pipeline significantly outperforms baseline methods in 3D reconstruction quality, paving the way for more adept robotic perception and interaction with transparent objects.
Unsupervised depth completion and estimation methods are trained by minimizing reconstruction error. Block artifacts from resampling, intensity saturation, and occlusions are amongst the many undesirable by-products of common data augmentation schemes that affect image reconstruction quality, and thus the training signal. Hence, typical augmentations on images viewed as essential to training pipelines in other vision tasks have seen limited use beyond small image intensity changes and flipping. The sparse depth modality in depth completion have seen even less use as intensity transformations alter the scale of the 3D scene, and geometric transformations may decimate the sparse points during resampling. We propose a method that unlocks a wide range of previously-infeasible geometric augmentations for unsupervised depth completion and estimation. This is achieved by reversing, or ``undo''-ing, geometric transformations to the coordinates of the output depth, warping the depth map back to the original reference frame. This enables computing the reconstruction losses using the original images and sparse depth maps, eliminating the pitfalls of naive loss computation on the augmented inputs and allowing us to scale up augmentations to boost performance. We demonstrate our method on indoor (VOID) and outdoor (KITTI) datasets, where we consistently improve upon recent methods across both datasets as well as generalization to four other datasets. Code available at: https://github.com/alexklwong/augundo.
Existing deep learning-based depth completion methods generally employ massive stacked layers to predict the dense depth map from sparse input data. Although such approaches greatly advance this task, their accompanied huge computational complexity hinders their practical applications. To accomplish depth completion more efficiently, we propose a novel lightweight deep network framework, the Long-short Range Recurrent Updating (LRRU) network. Without learning complex feature representations, LRRU first roughly fills the sparse input to obtain an initial dense depth map, and then iteratively updates it through learned spatially-variant kernels. Our iterative update process is content-adaptive and highly flexible, where the kernel weights are learned by jointly considering the guidance RGB images and the depth map to be updated, and large-to-small kernel scopes are dynamically adjusted to capture long-to-short range dependencies. Our initial depth map has coarse but complete scene depth information, which helps relieve the burden of directly regressing the dense depth from sparse ones, while our proposed method can effectively refine it to an accurate depth map with less learnable parameters and inference time. Experimental results demonstrate that our proposed LRRU variants achieve state-of-the-art performance across different parameter regimes. In particular, the LRRU-Base model outperforms competing approaches on the NYUv2 dataset, and ranks 1st on the KITTI depth completion benchmark at the time of submission. Project page: https://npucvr.github.io/LRRU/.
Depth perception models are typically trained on non-interactive datasets with predefined camera trajectories. However, this often introduces systematic biases into the learning process correlated to specific camera paths chosen during data acquisition. In this paper, we investigate the role of how data is collected for learning depth completion, from a robot navigation perspective, by leveraging 3D interactive environments. First, we evaluate four depth completion models trained on data collected using conventional navigation techniques. Our key insight is that existing exploration paradigms do not necessarily provide task-specific data points to achieve competent unsupervised depth completion learning. We then find that data collected with respect to photometric reconstruction has a direct positive influence on model performance. As a result, we develop an active, task-informed, depth uncertainty-based motion planning approach for learning depth completion, which we call DEpth Uncertainty-guided eXploration (DEUX). Training with data collected by our approach improves depth completion by an average greater than 18% across four depth completion models compared to existing exploration methods on the MP3D test set. We show that our approach further improves zero-shot generalization, while offering new insights into integrating robot learning-based depth estimation.
3D object detection at long range is crucial for ensuring the safety and efficiency of self driving vehicles, allowing them to accurately perceive and react to objects, obstacles, and potential hazards from a distance. But most current state of the art LiDAR based methods are range limited due to sparsity at long range, which generates a form of domain gap between points closer to and farther away from the ego vehicle. Another related problem is the label imbalance for faraway objects, which inhibits the performance of Deep Neural Networks at long range. To address the above limitations, we investigate two ways to improve long range performance of current LiDAR based 3D detectors. First, we combine two 3D detection networks, referred to as range experts, one specializing at near to mid range objects, and one at long range 3D detection. To train a detector at long range under a scarce label regime, we further weigh the loss according to the labelled point's distance from ego vehicle. Second, we augment LiDAR scans with virtual points generated using Multimodal Virtual Points (MVP), a readily available image-based depth completion algorithm. Our experiments on the long range Argoverse2 (AV2) dataset indicate that MVP is more effective in improving long range performance, while maintaining a straightforward implementation. On the other hand, the range experts offer a computationally efficient and simpler alternative, avoiding dependency on image-based segmentation networks and perfect camera-LiDAR calibration.
Depth completion is a popular research direction in the field of depth estimation. The fusion of color and depth features is the current critical challenge in this task, mainly due to the asymmetry between the rich scene details in color images and the sparse pixels in depth maps. To tackle this issue, we design an efficient Gated Cross-Attention Network that propagates confidence via a gating mechanism, simultaneously extracting and refining key information in both color and depth branches to achieve local spatial feature fusion. Additionally, we employ an attention network based on the Transformer in low-dimensional space to effectively fuse global features and increase the network's receptive field. With a simple yet efficient gating mechanism, our proposed method achieves fast and accurate depth completion without the need for additional branches or post-processing steps. At the same time, we use the Ray Tune mechanism with the AsyncHyperBandScheduler scheduler and the HyperOptSearch algorithm to automatically search for the optimal number of module iterations, which also allows us to achieve performance comparable to state-of-the-art methods. We conduct experiments on both indoor and outdoor scene datasets. Our fast network achieves Pareto-optimal solutions in terms of time and accuracy, and at the time of submission, our accurate network ranks first among all published papers on the KITTI official website in terms of accuracy.
We present NeuralLabeling, a labeling approach and toolset for annotating 3D scenes using either bounding boxes or meshes and generating segmentation masks, affordance maps, 2D bounding boxes, 3D bounding boxes, 6DOF object poses, depth maps, and object meshes. NeuralLabeling uses Neural Radiance Fields (NeRF) as a renderer, allowing labeling to be performed using 3D spatial tools while incorporating geometric clues such as occlusions, relying only on images captured from multiple viewpoints as input. To demonstrate the applicability of NeuralLabeling to a practical problem in robotics, we added ground truth depth maps to 30000 frames of transparent object RGB and noisy depth maps of glasses placed in a dishwasher captured using an RGBD sensor, yielding the Dishwasher30k dataset. We show that training a simple deep neural network with supervision using the annotated depth maps yields a higher reconstruction performance than training with the previously applied weakly supervised approach. We also show how instance segmentation and depth completion datasets generated using NeuralLabeling can be incorporated into a robot application for grasping transparent objects placed in a dishwasher with an accuracy of 83.3%, compared to 16.3% without depth completion.
Depth completion aims to recover dense depth maps from sparse ones, where color images are often used to facilitate this task. Recent depth methods primarily focus on image guided learning frameworks. However, blurry guidance in the image and unclear structure in the depth still impede their performance. To tackle these challenges, we explore a repetitive design in our image guided network to gradually and sufficiently recover depth values. Specifically, the repetition is embodied in both the image guidance branch and depth generation branch. In the former branch, we design a dense repetitive hourglass network (DRHN) to extract discriminative image features of complex environments, which can provide powerful contextual instruction for depth prediction. In the latter branch, we present a repetitive guidance (RG) module based on dynamic convolution, in which an efficient convolution factorization is proposed to reduce the complexity while modeling high-frequency structures progressively. Furthermore, in the semantic guidance branch, we utilize the well-known large vision model, i.e., segment anything (SAM), to supply RG with semantic prior. In addition, we propose a region-aware spatial propagation network (RASPN) for further depth refinement based on the semantic prior constraint. Finally, we collect a new dataset termed TOFDC for the depth completion task, which is acquired by the time-of-flight (TOF) sensor and the color camera on smartphones. Extensive experiments demonstrate that our method achieves state-of-the-art performance on KITTI, NYUv2, Matterport3D, 3D60, VKITTI, and our TOFDC.
RGB-guided depth completion aims at predicting dense depth maps from sparse depth measurements and corresponding RGB images, where how to effectively and efficiently exploit the multi-modal information is a key issue. Guided dynamic filters, which generate spatially-variant depth-wise separable convolutional filters from RGB features to guide depth features, have been proven to be effective in this task. However, the dynamically generated filters require massive model parameters, computational costs and memory footprints when the number of feature channels is large. In this paper, we propose to decompose the guided dynamic filters into a spatially-shared component multiplied by content-adaptive adaptors at each spatial location. Based on the proposed idea, we introduce two decomposition schemes A and B, which decompose the filters by splitting the filter structure and using spatial-wise attention, respectively. The decomposed filters not only maintain the favorable properties of guided dynamic filters as being content-dependent and spatially-variant, but also reduce model parameters and hardware costs, as the learned adaptors are decoupled with the number of feature channels. Extensive experimental results demonstrate that the methods using our schemes outperform state-of-the-art methods on the KITTI dataset, and rank 1st and 2nd on the KITTI benchmark at the time of submission. Meanwhile, they also achieve comparable performance on the NYUv2 dataset. In addition, our proposed methods are general and could be employed as plug-and-play feature fusion blocks in other multi-modal fusion tasks such as RGB-D salient object detection.
Dense SLAM based on monocular cameras does indeed have immense application value in the field of AR/VR, especially when it is performed on a mobile device. In this paper, we propose a novel method that integrates a light-weight depth completion network into a sparse SLAM system using a multi-basis depth representation, so that dense mapping can be performed online even on a mobile phone. Specifically, we present a specifically optimized multi-basis depth completion network, called BBC-Net, tailored to the characteristics of traditional sparse SLAM systems. BBC-Net can predict multiple balanced bases and a confidence map from a monocular image with sparse points generated by off-the-shelf keypoint-based SLAM systems. The final depth is a linear combination of predicted depth bases that can be optimized by tuning the corresponding weights. To seamlessly incorporate the weights into traditional SLAM optimization and ensure efficiency and robustness, we design a set of depth weight factors, which makes our network a versatile plug-in module, facilitating easy integration into various existing sparse SLAM systems and significantly enhancing global depth consistency through bundle adjustment. To verify the portability of our method, we integrate BBC-Net into two representative SLAM systems. The experimental results on various datasets show that the proposed method achieves better performance in monocular dense mapping than the state-of-the-art methods. We provide an online demo running on a mobile phone, which verifies the efficiency and mapping quality of the proposed method in real-world scenarios.
The reconstruction of textureless areas has long been a challenging problem in MVS due to lack of reliable pixel correspondences between images. In this paper, we propose the Textureless-aware Segmentation And Correlative Refinement guided Multi-View Stereo (TSAR-MVS), a novel method that effectively tackles challenges posed by textureless areas in 3D reconstruction through filtering, refinement and segmentation. First, we implement the joint hypothesis filtering, a technique that merges a confidence estimator with a disparity discontinuity detector to eliminate incorrect depth estimations. Second, to spread the pixels with confident depth, we introduce an iterative correlation refinement strategy that leverages RANSAC to generate 3D planes based on superpixels, succeeded by a weighted median filter for broadening the influence of accurately determined pixels. Finally, we present a textureless-aware segmentation method that leverages edge detection and line detection for accurately identify large textureless regions for further depth completion. Experiments on ETH3D, Tanks & Temples and Strecha datasets demonstrate the superior performance and strong generalization capability of our proposed method.
Depth completion, which aims to generate high-quality dense depth maps from sparse depth maps, has attracted increasing attention in recent years. Previous work usually employs RGB images as guidance, and introduces iterative spatial propagation to refine estimated coarse depth maps. However, most of the propagation refinement methods require several iterations and suffer from a fixed receptive field, which may contain irrelevant and useless information with very sparse input. In this paper, we address these two challenges simultaneously by revisiting the idea of deformable convolution. We propose an effective architecture that leverages deformable kernel convolution as a single-pass refinement module, and empirically demonstrate its superiority. To better understand the function of deformable convolution and exploit it for depth completion, we further systematically investigate a variety of representative strategies. Our study reveals that, different from prior work, deformable convolution needs to be applied on an estimated depth map with a relatively high density for better performance. We evaluate our model on the large-scale KITTI dataset and achieve state-of-the-art level performance in both accuracy and inference speed. Our code is available at https://github.com/AlexSunNik/ReDC.
Transparent objects present multiple distinct challenges to visual perception systems. First, their lack of distinguishing visual features makes transparent objects harder to detect and localize than opaque objects. Even humans find certain transparent surfaces with little specular reflection or refraction, like glass doors, difficult to perceive. A second challenge is that depth sensors typically used for opaque object perception cannot obtain accurate depth measurements on transparent surfaces due to their unique reflective properties. Stemming from these challenges, we observe that transparent object instances within the same category, such as cups, look more similar to each other than to ordinary opaque objects of that same category. Given this observation, the present paper explores the possibility of category-level transparent object pose estimation rather than instance-level pose estimation. We propose \textit{\textbf{TransNet}}, a two-stage pipeline that estimates category-level transparent object pose using localized depth completion and surface normal estimation. TransNet is evaluated in terms of pose estimation accuracy on a large-scale transparent object dataset and compared to a state-of-the-art category-level pose estimation approach. Results from this comparison demonstrate that TransNet achieves improved pose estimation accuracy on transparent objects. Moreover, we use TransNet to build an autonomous transparent object manipulation system for robotic pick-and-place and pouring tasks.
Depth completion is crucial for many robotic tasks such as autonomous driving, 3-D reconstruction, and manipulation. Despite the significant progress, existing methods remain computationally intensive and often fail to meet the real-time requirements of low-power robotic platforms. Additionally, most methods are designed for opaque objects and struggle with transparent objects due to the special properties of reflection and refraction. To address these challenges, we propose a Fast Depth Completion framework for Transparent objects (FDCT), which also benefits downstream tasks like object pose estimation. To leverage local information and avoid overfitting issues when integrating it with global information, we design a new fusion branch and shortcuts to exploit low-level features and a loss function to suppress overfitting. This results in an accurate and user-friendly depth rectification framework which can recover dense depth estimation from RGB-D images alone. Extensive experiments demonstrate that FDCT can run about 70 FPS with a higher accuracy than the state-of-the-art methods. We also demonstrate that FDCT can improve pose estimation in object grasping tasks. The source code is available at https://github.com/Nonmy/FDCT